Chemotherapy is a crucial treatment method for breast cancer, effectively targeting cancer cells. However, it also brings challenges, especially due to its effects on healthy cells and the toxicities it can cause. Attaining Pathological Complete Response (pCR) is a key determinant for successful cure and extended patient lifespan. To enhance predictive capabilities for achieving pCR and estimating potential Relapse-free Survival (RFS) time, we leveraged several machine learning models. Our study utilized a dataset comprising 400 observations and 117 features, consisting of both clinical and MRI features. Logistic regression worked best on the dataset for PCR, while Random Forest performed the best for RFS. These findings contribute to the ongoing efforts in refining treatment decisions for breast cancer patients, aiming to minimize adverse effects and optimize therapeutic outcomes.
This repository contains the implementation and analysis of a breast cancer pCR and RFS prediction models. The models are designed to enhance the accuracy of attaining pCR using various machine learning techniques.
The project focuses on predicting whether a pCR can be attained or not using the Wisconsin Diagnostic Breast Cancer (WDBC) dataset. The dataset includes features computed from digitized images of fine needle aspirates (FNA) of breast masses.
- Support Vector Machine (SVM)
- K-Nearest Neighbors (KNN)
- Decision Trees
The performance of each algorithm is thoroughly evaluated, with an emphasis on accuracy, precision, and recall.
This project demonstrates the potential of machine learning in medical diagnosis, particularly in enhancing the early detection of breast cancer.